Global and local batch process fault detection method based on dynamic orthogonality

Abstract

The invention provides a global and local batch process fault detection method based on dynamic orthogonality, which comprises the following steps of: (1) collecting each key variable data when a batch process is normally operated, and forming a training sample X which is an element of a set of RI*J*K in a normal operation state; (2) firstly expanding the training sample X into two-dimensional data X which is an element of a set of RI*KJ along a batch direction, then carrying out standardization on the expanded two-dimensional data, and rearranging the standardizated two-dimensional data intoX which is an element of a set of RKI*J; (3) on the basis of the two-dimensional data X which is the element of the set of RKI*J, establishing a time lag matrix XD to eliminate timing autocorrelationof a process variable; (4) constructing a dynamic orthogonal global local model for the established time lag matrix XD; (5) respectively establishing T2 and SPE statistical models in a characteristicspace and a residual space, and gaining a control limit; (6) collecting online process data and carrying out standardization processing; and (7) carrying out projection on the online data by utilizingthe established dynamic orthogonal global local model, and by the T2 and SPE statistical models, judging occurrence of a fault.

Description

technical field [0001] The invention belongs to the technical field of industrial process monitoring and relates to a dynamic orthogonal-based global and local intermittent process fault detection method. Background technique [0002] With the continuous expansion of the scale and complexity of modern industrial processes, batch process as an important production method is widely used in high-quality and small-capacity process production, such as special chemical industry, food production, pharmaceutical production and semiconductor production Wait. Batch process production is composed of many production links, each link is crucial to the final product quality and safe production, and the completion of the previous link directly affects the implementation of the next link. Batch process production is more prone to failure due to the nesting of multiple stages, increasing production scale and increasing complexity. If the failure cannot be detected and eliminated in time, t...

AI Technical Summary

Problems solved by technology

The traditional method only considers the global structure of the process data and ignores the local structure of the data during feature extraction, so the global and local structure of the process data cannot be fully maintained in the process of dimensionality reduction, and it is not considered in the data reconstruction. The generalized orthogonal problem causes the loss of some useful information in the process monitoring, resulting in poor monitoring effect

Images